CN114715146A - Method for predicting severity of potential collision accident - Google Patents
Method for predicting severity of potential collision accident Download PDFInfo
- Publication number
- CN114715146A CN114715146A CN202210497607.XA CN202210497607A CN114715146A CN 114715146 A CN114715146 A CN 114715146A CN 202210497607 A CN202210497607 A CN 202210497607A CN 114715146 A CN114715146 A CN 114715146A
- Authority
- CN
- China
- Prior art keywords
- collision
- vehicle
- accident
- severity
- collision accident
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000012512 characterization method Methods 0.000 claims abstract description 36
- 238000010276 construction Methods 0.000 claims abstract description 24
- 238000012216 screening Methods 0.000 claims abstract description 9
- 238000005070 sampling Methods 0.000 claims abstract description 8
- 238000012549 training Methods 0.000 claims description 6
- 238000013459 approach Methods 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000010219 correlation analysis Methods 0.000 claims description 3
- 238000011056 performance test Methods 0.000 claims description 3
- 238000010845 search algorithm Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 description 5
- 230000002265 prevention Effects 0.000 description 3
- 230000000875 corresponding effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000001604 Rao's score test Methods 0.000 description 1
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W30/00—Purposes of road vehicle drive control systems not related to the control of a particular sub-unit, e.g. of systems using conjoint control of vehicle sub-units
- B60W30/08—Active safety systems predicting or avoiding probable or impending collision or attempting to minimise its consequences
- B60W30/095—Predicting travel path or likelihood of collision
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W30/00—Purposes of road vehicle drive control systems not related to the control of a particular sub-unit, e.g. of systems using conjoint control of vehicle sub-units
- B60W30/08—Active safety systems predicting or avoiding probable or impending collision or attempting to minimise its consequences
- B60W30/095—Predicting travel path or likelihood of collision
- B60W30/0953—Predicting travel path or likelihood of collision the prediction being responsive to vehicle dynamic parameters
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
- B60W40/10—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models related to vehicle motion
- B60W40/105—Speed
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
- B60W40/12—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models related to parameters of the vehicle itself, e.g. tyre models
- B60W40/13—Load or weight
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W2520/00—Input parameters relating to overall vehicle dynamics
- B60W2520/06—Direction of travel
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W2520/00—Input parameters relating to overall vehicle dynamics
- B60W2520/10—Longitudinal speed
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W2530/00—Input parameters relating to vehicle conditions or values, not covered by groups B60W2510/00 or B60W2520/00
- B60W2530/10—Weight
Landscapes
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Transportation (AREA)
- Mechanical Engineering (AREA)
- Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Traffic Control Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for predicting the severity of a potential collision accident, which comprises the following steps: acquiring a collision accident sample, wherein a collision event in the collision accident sample is provided with a collision severity label, and the collision accident sample is divided into a plurality of collision accident subsamples based on the collision severity label; performing feature construction based on the basic features of the collision accident sample to obtain construction features, and screening the construction features to obtain characterization features relevant to the collision severity label; based on the characterization features, performing over-sampling or under-sampling treatment on the collision accident subsample to obtain a balanced collision accident sample; establishing a collision severity prediction model based on the obtained balanced collision accident sample; and predicting the severity of the potential collision accident based on the collision severity prediction model. The method can predict the severity of the potential collision accident in real time in the driving process before the collision accident occurs.
Description
Technical Field
The invention belongs to the technical field of traffic safety research, and particularly relates to a method for predicting the severity of a potential collision accident.
Background
Collision events occur frequently, and the evaluation of accident consequences is particularly important, so that a driver can know the severity of risks and take preventive measures in time. The severity of the previous collision accident is mostly measured by the actual loss after the accident, which is not favorable for the early prevention of the collision accident.
Therefore, a method capable of predicting the severity of a potential collision accident in real time is urgently needed, the severity of the accident possibly caused by collision in the current state is predicted, the consequences of the potential accident can be known in advance before the accident occurs, and early collision warning and accident prevention are facilitated.
Disclosure of Invention
Aiming at the problems, the invention provides a method for predicting the severity of a potential collision accident, which can predict the severity of the potential collision accident in real time based on the running state parameters of a vehicle. The method constructs a characteristic capable of representing the severity of the accident after the collision from the perspective of the energy change in the collision process based on the running state parameters of the vehicle when no collision occurs.
The invention provides a method for predicting the severity of a potential collision accident, which comprises the following steps: acquiring a collision accident sample, wherein a collision event in the collision accident sample is provided with a collision severity label, and the collision accident sample is divided into a plurality of collision accident subsamples based on the collision severity label; performing characteristic construction based on the basic characteristics of the collision accident sample to obtain construction characteristics, and screening the construction characteristics to obtain characterization characteristics related to the collision accident severity label; performing over-sampling or under-sampling treatment on the collision accident subsample based on the characterization feature to obtain a balanced collision accident sample; establishing a collision severity prediction model based on the obtained balanced collision accident sample; and predicting the severity of the potential collision accident based on the collision severity prediction model.
Preferably, the acquiring of the collision accident sample specifically includes: based on the collision type, a collision accident sample is obtained. According to whether the driving directions of the vehicles are the same or not, the collision can be divided into a same-direction collision and a reverse collision, the collision types are different, and the subsequently screened characterization features are slightly different, so that in order to improve the prediction effect, it is further preferable that the collision events related in the collision accident sample are the same-direction collision or the same reverse collision.
Preferably, the collision event involved in the collision accident sample is a collision event of two vehicles colliding with each other; the basic features of the crash event sample include: absolute speed of the vehicle involved in the accident, longitudinal speed of the vehicle involved in the accident, transverse speed of the vehicle involved in the accident, mass of the vehicle involved in the accident, and course angle of the vehicle involved in the accident; the construction features include: the vehicle speed control method comprises the following steps of changing the speed of the vehicle before and after collision, changing the longitudinal speed of the vehicle before and after collision, changing the lateral speed of the vehicle before and after collision, relative speed of the two vehicles and relative collision angle.
Preferably, the vehicles involved are numbered as i, i ═ 1 or 2; based on the basic characteristics of the collision accident sample, performing characteristic construction to obtain construction characteristics, wherein the method specifically comprises the following steps: amount of speed change before and after i-collision of vehicle involved in accident
Longitudinal vehicle speed variation before and after I collision of accident-related vehicle
Amount of change in lateral vehicle speed before and after i-collision of accident-related vehicle
Relative velocity of two vehicles Relative to abs (V)i-v3-i) Relative collision angle Relative- θ abs (θ)i-θ3-i) Wherein v isiFor absolute speed of the vehicle i involved in the accident，v3-iFor the absolute speed of the vehicle 3-i involved, vxiFor the longitudinal speed of the vehicle i involved, vx3-iFor the longitudinal speed of the vehicle 3-i involved, vyiFor the transverse speed of the vehicle i involved, vy3-iFor transverse speed of the vehicle 3-i involved, miFor mass of the vehicle i involved, m3-iFor mass related to the vehicle 3-i, abs refers to the absolute value function, θiFor the course angle, θ, of the vehicle i involved therein3-iFor the heading angle of the affected vehicle 3-i, α is the angle at which the two vehicles approach each other, α ═ θi+θ3-i。
Preferably, the screening the structural characteristics to obtain the characterization characteristics having correlation with the crash severity label further comprises: calculating the importance of the structural features through the Gini coefficient, and selecting the structural features with high importance; analyzing the correlation of the structural features through a Pearson correlation coefficient, and deleting redundant features; and screening out characteristic features from the structural features according to the results of the importance analysis and the correlation analysis.
Preferably, based on the characterization features, the oversampling or undersampling processing is performed on the collision accident subsample to obtain a balanced collision accident sample, which specifically includes: when the collision accident subsamples are a plurality of types of samples, deleting the intra-class outliers in the collision accident subsamples to realize undersampling of the collision accident subsamples; when the collision accident subsample is a minority class sample, oversampling the collision accident subsample based on the core seed cluster of the collision accident subsample.
Preferably, the method for predicting the severity of a collision is established based on the obtained balanced collision accident sample, and specifically includes: and obtaining a training sample based on the balanced collision accident sample, taking the characterization feature as input, taking the collision accident severity as output, and establishing a collision severity prediction model based on an XGboost algorithm.
Preferably, the method for predicting the severity of a collision is based on the obtained balanced collision accident sample, and further comprises: utilizing a grid search algorithm to carry out parameter adjustment on the collision severity prediction model; and/or performing performance test on the collision severity prediction model based on one or more indexes of accuracy, call, precision, f1_ score.
Preferably, the predicting the severity of the potential collision accident based on the collision severity prediction model specifically includes: acquiring real-time basic characteristics of the own vehicle and other vehicles according to the basic characteristics of the collision accident sample; performing feature construction based on the real-time basic features by referring to the method for acquiring the characterization features to obtain online characterization features; and inputting the obtained online characterization characteristics into the collision severity prediction model to obtain the real-time severity of the potential collision accident.
Preferably, the real-time basic characteristics of the self vehicle and other vehicles are acquired, and the method specifically comprises the following steps: acquiring real-time absolute speed of the self-vehicle and real-time relative speed of the self-vehicle and other vehicles, and acquiring real-time absolute speed of other vehicles according to the real-time absolute speed of the self-vehicle and the real-time relative speed of the self-vehicle and other vehicles; acquiring real-time position parameters of the self-vehicle and other vehicles, and acquiring real-time course angles of the self-vehicle and other vehicles according to the real-time position parameters; and obtaining the vehicle types of the self vehicle and other vehicles, and obtaining the quality of the self vehicle and other vehicles according to the vehicle types.
Compared with the prior art, the invention has the following beneficial effects:
(1) the method carries out feature construction and screening based on the basic features of the accident-related vehicles such as speed, quality, course angle and the like to obtain the characterization features which are relevant to the severity of the collision accident, and the characterization features are used as the input features of the training collision severity prediction model, so that the prediction effect of the obtained prediction model is better.
(2) By using the prediction model and the method, the severity grade possibly caused by the collision accident at the current moment of the vehicle can be predicted only based on the running state parameters (namely the real-time basic characteristics) of the vehicle involved in the accident at the current moment, the real-time performance is high, and the prediction model and the prediction method have important significance in driving early warning and traffic accident prevention.
Drawings
FIG. 1 is a schematic flow chart illustrating a method for predicting the severity of a potential collision accident according to embodiment 1 of the present invention;
fig. 2 is a graph showing the fitting effect of the collision severity prediction model in embodiment 1 of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely below, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
The embodiment provides a method for predicting the severity of a potential collision accident, which comprises the following steps:
(1) acquiring a collision accident sample, wherein collision events in the collision accident sample are provided with collision severity labels, and the collision accident sample is divided into a plurality of collision accident subsamples based on the collision severity labels.
The collision accident sample can be collected by self, and can also be obtained by utilizing a third-party website or a database and the like. In the embodiment, as a preferred scheme, the collision accident sample is obtained from a NHTSA collision sample library in the united states, which is an open-source database, and the collision sample library contains abundant collision cases and a large amount of information, and all collision events are labeled with collision severity. The crash severity labels severity of the crash events in the crash accident samples obtained in this embodiment may be divided into three categories, which are respectively the severity1 representing a light crash, the severity2 representing a medium crash, and the severity3 representing a heavy crash, and based on the crash severity labels, the crash accident samples may be divided into three crash accident sub-samples, which are respectively the light crash accident sub-sample crash1, the medium crash accident sub-sample crash2, and the heavy crash accident sub-sample crash 3.
In the embodiment, as a preferable scheme, a collision accident sample is obtained based on collision types, and in consideration of that the characterization characteristics corresponding to collision events of different collision types are slightly different, in order to establish a more accurate and better-effect prediction model, the collision events related to the collision accident sample are collision events of the same collision type, both of which are co-directional collisions or both of which are reverse collisions, wherein the co-directional collisions and the reverse collisions further include vehicle collisions in the same lane (rear-end collisions in the co-directional collisions or front-end collisions in the reverse collisions) and vehicle collisions in adjacent lanes (side collisions in the same driving direction or side collisions in opposite driving directions). For example, in the present embodiment, a prediction model is established for a research scenario of a same-direction multi-lane, and all collision events involved in the acquired collision accident sample are same-direction collisions, including a rear-end collision and a side collision in the same driving direction.
(2) And performing characteristic construction based on the basic characteristics of the collision accident sample to obtain construction characteristics, and screening the construction characteristics to obtain characterization characteristics related to the collision severity label.
In this embodiment, as a preferable scheme, the collision event related to the collision accident sample is a collision event of two vehicles colliding with each other, and the number of the collision-related vehicles is denoted as i, i is 1 or 2, where 1 is a leading vehicle and 2 is a trailing vehicle.
Among the factors affecting the severity of the crash, the factors related to the vehicles involved in the crash mainly include the speed, mass and impact location of the vehicles, and in order to predict the severity of the crash, the basic features corresponding to the factors in the crash sample are first analyzed. As will be appreciated by those skilled in the art, in studying the severity of a crash event, the basic characteristics of the crash event sample refer to the vehicle base parameters at the time of the crash event, such as the crash event being t0-t1The time period and the collision occurrence time are t0The time of day.
The basic features of the crash event sample include: absolute speed v of the vehicle concerned (including the absolute speed v of the vehicle concerned 1)1And the absolute speed v of the vehicle 2 involved2) Longitudinal speed vx of the vehicle involved (including the longitudinal speed vx of the vehicle involved 1)1And the longitudinal speed vx of the affected vehicle 22) The lateral speed vy of the vehicle involved in the accident (including the lateral speed vy of the vehicle 1 involved in the accident1And the transverse speed vy of the affected vehicle 22) Mass m of the vehicle involved (including mass m of the vehicle involved 1)1And mass m of the accident vehicle 22) Heading angle theta of the vehicle concerned (including heading angle theta of the vehicle concerned 1)1And the heading angle theta of the affected vehicle 22)。
For the crash accident sample of the present embodiment, the correlation between the basic features and the crash severity label reliability is analyzed, and is characterized by the pearson correlation coefficient r as shown in table 1:
TABLE 1 Pearson correlation coefficient between base signature and crash severity label
| r | v1 | vx1 | vy1 | v2 | vx2 | vy2 | m1 | m2 | θ1 | θ2 |
| severity | 0.127 | 0.115 | 0.071 | -0.017 | -0.011 | 0.009 | -0.189 | 0.062 | -0.018 | -0.032 |
It is generally considered that the absolute value of the pearson correlation coefficient r is strong correlation between the two at 0.8 or more, weak correlation between the two at 0.3 to 0.8, and no correlation between the two at 0.3 or less.
As can be seen from table 1, there is no correlation between each basic feature of the crash accident sample and the crash severity label, and the potential crash accident severity of the vehicle cannot be predicted through the basic features. Therefore, the severity degree of the collision is not influenced by the mass, the speed and the collision position independently, but influenced by the collision severity degree together, and the purpose of predicting the severity degree of the potential collision accident cannot be realized only by basic characteristics.
In order to obtain the characterization feature having correlation with the crash severity label so as to be able to predict the severity of the potential crash accident, the present embodiment performs the feature construction based on the above-described basic features. When the vehicle collides, energy transfer occurs, i.e., kinetic energy is converted into other energy, which causes vehicle deformation and personnel injury. Thus, the present implementation takes into account the variation in kinetic energy, and selects the following features for construction, including: the speed variation DeltaV before and after the collision of the vehicle (including the speed variation DeltaV before and after the collision of the vehicle 1)1Speed variation DeltaV before and after collision with the accident vehicle 22) The longitudinal vehicle speed variation DeltaVx before and after the collision of the vehicle (including the longitudinal vehicle speed variation DeltaVx before and after the collision of the vehicle 1)1Longitudinal vehicle speed variation DeltaVx before and after collision with the accident-related vehicle 22) The amount of change DeltaVy in the lateral vehicle speed before and after the collision with the vehicle (including the amount of change DeltaVy in the lateral vehicle speed before and after the collision with the vehicle 1)1The amount of change DeltaVy in the lateral speed before and after the collision with the vehicle 22) The Relative speed of both vehicles Relative to each other-V, and the Relative collision angle Relative _ theta.
However, it is difficult to intuitively acquire the change process of the basic parameters of the vehicle because the time of the collision process is short. Therefore, the present embodiment describes the collision process of the vehicle from the viewpoint of conservation of momentum during collision when performing the feature configuration, which is performed based on the above-described basic features, and constructs a feature that can characterize the severity of the accident after collision. The method specifically comprises the following steps:
amount of speed change before and after i-collision of vehicle involved in accident
Longitudinal vehicle speed variation before and after I collision of accident-related vehicle
Amount of change in lateral vehicle speed before and after i-collision of accident-related vehicle
Relative velocity of two vehicles Relative to abs (V)i-v3-i)，
Relative collision angle Relative- θ abs (θ)i-θ3-i)，
Wherein v isiTo the absolute speed of the vehicle i involved, v3-iTo relate to the absolute speed of the vehicle 3-i,
vxifor longitudinal speed of the vehicle i involved, vx3-iFor longitudinal speed of the vehicle 3-i involved in the eventThe degree of the magnetic field is measured,
vyifor the transverse speed of the vehicle i involved, vy3-iTo relate to the lateral speed of the vehicle 3-i,
mifor mass of the vehicle i involved, m3-iTo relate to the mass of the vehicle 3-i,
abs refers to the function of the absolute value,
θifor the course angle, θ, of the vehicle i involved therein3-iFor the heading angle of the vehicle of interest 3-i,
alpha is the angle at which the two vehicles approach each other, and alpha is thetai+θ3-i。
For the crash accident sample of the present embodiment, the correlation between the above-mentioned structural characteristics and the crash severity label safety is analyzed, and is characterized by the pearson correlation coefficient r as shown in table 2:
TABLE 2 Pearson correlation coefficient between build characteristics and crash severity labels
| r | DeltaV1 | DeltaV2 | DeltaVx1 | DeltaVx2 | DeltaVy1 | DeltaVy2 | Relative-V | Relative_θ |
| severity | 0.877 | 0.695 | 0.872 | 0.696 | 0.125 | 0.097 | 0.401 | -0.237 |
As can be seen from table 2, there are features having a correlation with the collision severity label in the structural features, and the structural features having a correlation with the collision severity label are screened out as the characterization features.
As a preferable scheme, in order to reduce the complexity of the model, find a characterization feature more suitable for characterizing the severity of the accident, and screen the structural features to obtain a characterization feature correlated with the collision severity label, the method further includes: calculating the importance of the structural features through the Gini coefficient, and selecting the structural features with high importance; analyzing the correlation of the structural features through a Pearson correlation coefficient, and deleting redundant features; and screening out characteristic features from the structural features according to the results of the importance analysis and the correlation analysis. Further details regarding this step are described in the patent application No. 2022101760035, which precedes this unit and is not described herein. The acquired collision accident sample is screened for structural characteristics, and finally the characteristic that the severity of the collision accident is characterized by DeltaV1And Relative-V. If the acquired collision accident sample is reverse collision, the method of the characteristic structure is the same as the embodiment, the obtained structural characteristics are the same as the embodiment, but the characteristic is screened outWhen in characteristic, because the importance degree of the structural characteristics is ordered slightly differently, the finally obtained characteristic characteristics may be different from the collision accident sample of the same-direction collision.
(3) And performing over-sampling or under-sampling treatment on the collision accident subsample based on the characterization characteristics to obtain a balanced collision accident sample.
In the field of accident security, there is a common and real phenomenon: collision events with high severity are rare, while slight collision events are common, which causes the phenomenon that the number of collision samples with different severity levels is unbalanced, and the training of a subsequent collision severity prediction model is influenced, so that the prediction result is biased to the severity level with more samples. Therefore, it needs to be equalized and then used.
As a preferred scheme, the present embodiment specifically includes: when the collision accident subsamples are a plurality of types of samples, deleting the intra-class outliers in the collision accident subsamples to realize undersampling of the collision accident subsamples; when the collision accident subsample is a minority class sample, oversampling the collision accident subsample based on the core seed cluster of the collision accident subsample. Further details of the method are described in the patent application No. 2022101760035, which is filed earlier in this section, and are not repeated herein. Of course, other processing methods in the prior art can be used to perform the equalization processing on the collision accident sample.
(4) And establishing a collision severity prediction model based on the obtained balanced collision accident sample.
In this embodiment, as a preferred scheme, a training sample is obtained based on the equalized collision accident sample, the characterization feature is used as an input, the collision accident severity is used as an output, and a collision severity prediction model is established based on an XGBoost algorithm; and utilizing a grid search algorithm to perform parameter adjustment on the collision severity prediction model, and performing performance test on the collision severity prediction model based on one or more indexes of accuracy, call, precision and f1_ score. The method comprises the following specific steps:
the equalized collision accident sample is divided into a training sample (train) and a test sample (test). The input adopts the characterization feature group of the accident severity screened out above: { Deltav1, relative _ v }, the output is the severity of the crash: { criterion 1, criterion 2, criterion 3 }. And constructing an initial XGboost multi-classification model in python by utilizing an XGboost toolkit according to default parameters. In order to improve the model effect, the hyper-parameters of the model are adjusted by using a grid search (GridSearchCV) algorithm.
Based on the performance of common classification indexes accuracy, call, precision and f1_ score test models, aiming at the balanced collision accident sample obtained in the embodiment, a collision severity prediction model is established based on the XGboost algorithm, the model fitting effect is shown in FIG. 2, and the model performance is shown in Table 3:
TABLE 3 model Performance
| accuracy | precision | recall | f1_score | |
| Model performance | 0.9068 | 0.9078 | 0.9074 | 0.9074 |
The requirements on accuracy and timeliness are high in the prediction of the severity of the collision accident, and as can be seen from table 3 and fig. 2, a prediction model obtained based on the XGboost algorithm is excellent in performance.
(5) And predicting the severity of the potential collision accident based on the collision severity prediction model.
Firstly, acquiring real-time basic characteristics of the own vehicle and other vehicles, namely acquiring real-time absolute speed v of the own vehicle by referring to the basic characteristics of the collision accident sample1', real-time absolute speed v of other vehicle2' real-time longitudinal speed of self vehicle vx1' real-time longitudinal speed of other vehicle vx2' real-time transverse speed vy of bicycle2' real-time transverse speed vy of other vehicles2' mass m of bicycle1' mass m of other vehicle2' real-time course angle theta of own vehicle1', heading angle of other vehicle theta2’。
Wherein the real-time absolute speed v of the vehicle1The real-time relative speed of the vehicle and other vehicles can be measured by a sensor and other devices, and the real-time absolute speed v of the vehicle can be directly read1'and real-time relative speed of the own vehicle and the other vehicle relative _ v' can obtain the real-time relative speed v of the other vehicle2’＝v1'-relative _ v'; setting a reference coordinate system, namely obtaining a real-time longitudinal speed and a real-time transverse speed according to the real-time absolute speed; mass m of the vehicle1The vehicle type of the other vehicle can be directly obtained, and the vehicle type of the other vehicle can be obtained according to the vehicle-mounted camera, so that the quality estimation is carried out, and the quality m of the other vehicle is obtained2'. The position of the self-vehicle and other vehicles is obtained, and the position parameter (x) of the self-vehicle can be obtained according to the reference coordinate system1，y1) And position parameters (x) of other vehicles2，y2) According to the position parameter, the course angle of the self-vehicle can be obtained
Course angle of other vehicle
And secondly, performing feature construction based on the real-time basic features by referring to the method for acquiring the characterization features to obtain online characterization features.
For the prediction model of the embodiment, when the severity of the co-directional collision is predicted, the online characterization feature includes real-time relative speed relative _ v' of the own vehicle and other vehicles, and if the own vehicle is predicted to be the active vehicle, the online characterization feature is also required to be the speed variation before and after the collision of the own vehicle
If the other vehicle is predicted to be the active vehicle, the on-line characterization feature required is the speed variation before and after the collision of the other vehicle
Where α' is a real-time approach angle between the own vehicle and another vehicle, and α ═ θ1’+θ2'; real-time relative speed relative _ v ═ abs (v) of the own vehicle and the other vehicle1’-v2’)。
And thirdly, inputting the obtained online characterization characteristics into the collision severity prediction model to obtain the real-time severity of the potential collision accident.
The above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art may still modify the technical solutions described in the foregoing embodiments, or may equally substitute some or all of the technical features; such modifications and substitutions do not depart from the spirit and scope of the present invention, and they should be construed as being included in the following claims and description.
Claims (10)
1. A method for predicting the severity of a potential collision accident, comprising the steps of:
acquiring a collision accident sample, wherein a collision event in the collision accident sample is provided with a collision severity label, and the collision accident sample is divided into a plurality of collision accident subsamples based on the collision severity label;
performing characteristic construction based on the basic characteristics of the collision accident sample to obtain construction characteristics, and screening the construction characteristics to obtain characterization characteristics related to the collision severity label;
based on the characterization features, performing over-sampling or under-sampling treatment on the collision accident subsample to obtain a balanced collision accident sample;
establishing a collision severity prediction model based on the obtained balanced collision accident sample;
and predicting the severity of the potential collision accident based on the collision severity prediction model.
2. The method for predicting the severity of a potential collision accident according to claim 1, wherein obtaining a sample of the collision accident comprises: based on the collision type, a collision accident sample is obtained.
3. The method of predicting the severity of a potential collision accident according to claim 1 or 2, wherein the collision event involved in the sample of collision accidents is a collision event of a two-vehicle collision;
the basic features of the crash event sample include: absolute speed of the vehicle involved in the accident, longitudinal speed of the vehicle involved in the accident, transverse speed of the vehicle involved in the accident, mass of the vehicle involved in the accident, and course angle of the vehicle involved in the accident;
the construction features include: the vehicle speed change amount before and after the collision of the accident-related vehicle, the longitudinal vehicle speed change amount before and after the collision of the accident-related vehicle, the lateral vehicle speed change amount before and after the collision of the accident-related vehicle, the relative speed of the two vehicles and the relative collision angle.
4. The method of predicting the severity of a potential collision accident according to claim 3, wherein the vehicles involved in the accident are numbered as i, i-1 or 2; based on the basic characteristics of the collision accident sample, performing characteristic construction to obtain construction characteristics, wherein the method specifically comprises the following steps:
amount of speed change before and after i-collision of vehicle involved in accident
Longitudinal vehicle speed variation before and after I collision of accident-related vehicle
Amount of change in lateral vehicle speed before and after i-collision of accident-related vehicle
Relative velocity of two vehicles Relative to abs (V)i-v3-i)，
Relative collision angle Relative- θ abs (θ)i-θ3-i)，
Wherein v isiTo the absolute speed of the vehicle i involved, v3-iTo relate to the absolute speed of the vehicle 3-i,
vxifor longitudinal speed of the vehicle i involved, vx3-iTo relate to the longitudinal speed of the vehicle 3-i,
vyifor transverse speed of the vehicle i involved, vy3-iTo relate to the lateral speed of the vehicle 3-i,
mifor mass of the vehicle i involved, m3-iIn order to be involved with the mass of the vehicle 3-i,
θifor the course angle, θ, of the vehicle i involved therein3-iFor the heading angle of the vehicle of interest 3-i,
alpha is the angle at which the two vehicles approach each other, and alpha is thetai+θ3-i。
5. The method of predicting the severity of a potential collision accident according to claim 1, wherein the structural features are screened for characterization features having a correlation with the collision severity label, further comprising:
calculating the importance of the structural features through the Gini coefficient, and selecting the structural features with high importance;
analyzing the correlation of the structural features through a Pearson correlation coefficient, and deleting redundant features;
and screening out characteristic features from the structural features according to the results of the importance analysis and the correlation analysis.
6. The method according to claim 1, wherein the oversampling or undersampling is performed on the crash accident subsample based on the characterization feature to obtain a balanced crash accident sample, and specifically comprises:
when the collision accident subsamples are a plurality of types of samples, deleting the intra-class outliers in the collision accident subsamples to realize undersampling of the collision accident subsamples;
when the collision accident subsample is a minority class sample, oversampling the collision accident subsample based on the core seed cluster of the collision accident subsample.
7. The method for predicting the severity of a potential collision accident according to claim 1, wherein the step of establishing a prediction model of the severity of the collision based on the obtained equalized collision accident samples comprises:
and obtaining a training sample based on the balanced collision accident sample, taking the characterization feature as input, taking the collision accident severity as output, and establishing a collision severity prediction model based on an XGboost algorithm.
8. The method of predicting the severity of a potential collision accident according to claim 7, wherein the step of building a prediction model of the severity of the collision based on the resulting equalized sample of collision accidents further comprises: utilizing a grid search algorithm to perform parameter adjustment on the collision severity prediction model;
and/or performing performance test on the collision severity prediction model based on one or more indexes of accuracy, call, precision and f1_ score.
9. The method for predicting the severity of a potential collision accident according to claim 1, wherein predicting the severity of a potential collision accident based on the prediction model of the severity of a collision accident comprises:
acquiring real-time basic characteristics of the own vehicle and other vehicles according to the basic characteristics of the collision accident sample;
performing feature construction based on the real-time basic features by referring to the method for acquiring the characterization features to obtain online characterization features;
and inputting the obtained online characterization characteristics into the collision severity prediction model to obtain the real-time severity of the potential collision accident.
10. The method for predicting the severity of a potential collision accident according to claim 9, wherein the real-time basic characteristics of the own vehicle and the other vehicles are obtained, and the method specifically comprises the following steps:
acquiring the real-time absolute speed of the self-vehicle and the real-time relative speed of the self-vehicle and other vehicles, and acquiring the real-time absolute speed of other vehicles according to the real-time absolute speed of the self-vehicle and the real-time relative speed of the self-vehicle and other vehicles;
acquiring real-time position parameters of the self-vehicle and other vehicles, and acquiring real-time course angles of the self-vehicle and other vehicles according to the real-time position parameters;
and obtaining the vehicle types of the self vehicle and other vehicles, and obtaining the quality of the self vehicle and other vehicles according to the vehicle types.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210497607.XA CN114715146B (en) | 2022-05-09 | 2022-05-09 | Method for predicting severity of potential collision accident |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210497607.XA CN114715146B (en) | 2022-05-09 | 2022-05-09 | Method for predicting severity of potential collision accident |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114715146A true CN114715146A (en) | 2022-07-08 |
| CN114715146B CN114715146B (en) | 2024-08-16 |
Family
ID=82231993
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210497607.XA Active CN114715146B (en) | 2022-05-09 | 2022-05-09 | Method for predicting severity of potential collision accident |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114715146B (en) |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005103674A (en) * | 2003-09-29 | 2005-04-21 | Yaskawa Electric Corp | Apparatus and method for controlling robot |
| CN110033386A (en) * | 2019-03-07 | 2019-07-19 | 阿里巴巴集团控股有限公司 | Identification method and device, the electronic equipment of car accident |
| CN110276370A (en) * | 2019-05-05 | 2019-09-24 | 南京理工大学 | A kind of road traffic accident risk Factor Analysis method based on random forest |
| CN111582368A (en) * | 2020-05-07 | 2020-08-25 | 中南大学 | Comprehensive evaluation method for safety performance of passenger vehicle based on accident statistical data |
| CN112487617A (en) * | 2020-11-19 | 2021-03-12 | 平安科技（深圳）有限公司 | Collision model-based risk prevention method, device, equipment and storage medium |
| CN112966352A (en) * | 2021-03-10 | 2021-06-15 | 西华大学 | System and method for rapidly calculating vehicle deformation collision energy in traffic accident |
| US20210188263A1 (en) * | 2019-12-23 | 2021-06-24 | Baidu International Technology (Shenzhen) Co., Ltd. | Collision detection method, and device, as well as electronic device and storage medium |
| CN113094808A (en) * | 2021-03-22 | 2021-07-09 | 大连理工大学 | Simulation data and artificial intelligence based automobile collision damage grade real-time prediction method |
| CN113256993A (en) * | 2021-07-15 | 2021-08-13 | 杭州华鲤智能科技有限公司 | Method for training and analyzing vehicle driving risk by model |
| US20220118931A1 (en) * | 2020-10-21 | 2022-04-21 | Cambridge Mobile Telematics Inc. | Method and system for vehicle crash prediction using multi-vehicle data |
| CN114418034A (en) * | 2022-02-25 | 2022-04-29 | 吉林大学 | Processing method and system for class imbalance samples |
| TWM626569U (en) * | 2022-01-06 | 2022-05-01 | 國立清華大學 | car accident severity prediction device |
-
2022
- 2022-05-09 CN CN202210497607.XA patent/CN114715146B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005103674A (en) * | 2003-09-29 | 2005-04-21 | Yaskawa Electric Corp | Apparatus and method for controlling robot |
| CN110033386A (en) * | 2019-03-07 | 2019-07-19 | 阿里巴巴集团控股有限公司 | Identification method and device, the electronic equipment of car accident |
| CN110276370A (en) * | 2019-05-05 | 2019-09-24 | 南京理工大学 | A kind of road traffic accident risk Factor Analysis method based on random forest |
| US20210188263A1 (en) * | 2019-12-23 | 2021-06-24 | Baidu International Technology (Shenzhen) Co., Ltd. | Collision detection method, and device, as well as electronic device and storage medium |
| CN111582368A (en) * | 2020-05-07 | 2020-08-25 | 中南大学 | Comprehensive evaluation method for safety performance of passenger vehicle based on accident statistical data |
| US20220118931A1 (en) * | 2020-10-21 | 2022-04-21 | Cambridge Mobile Telematics Inc. | Method and system for vehicle crash prediction using multi-vehicle data |
| CN112487617A (en) * | 2020-11-19 | 2021-03-12 | 平安科技（深圳）有限公司 | Collision model-based risk prevention method, device, equipment and storage medium |
| CN112966352A (en) * | 2021-03-10 | 2021-06-15 | 西华大学 | System and method for rapidly calculating vehicle deformation collision energy in traffic accident |
| CN113094808A (en) * | 2021-03-22 | 2021-07-09 | 大连理工大学 | Simulation data and artificial intelligence based automobile collision damage grade real-time prediction method |
| CN113256993A (en) * | 2021-07-15 | 2021-08-13 | 杭州华鲤智能科技有限公司 | Method for training and analyzing vehicle driving risk by model |
| TWM626569U (en) * | 2022-01-06 | 2022-05-01 | 國立清華大學 | car accident severity prediction device |
| CN114418034A (en) * | 2022-02-25 | 2022-04-29 | 吉林大学 | Processing method and system for class imbalance samples |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114715146B (en) | 2024-08-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111832225B (en) | Method for constructing driving condition of automobile | |
| CN112487617B (en) | Collision model-based risk prevention method, device, equipment and storage medium | |
| CN106022561A (en) | Driving comprehensive evaluation method | |
| CN111243338A (en) | Vehicle acceleration-based collision risk evaluation method | |
| CN107705635B (en) | Method for judging traffic conflict of electric bicycles at signalized intersection | |
| CN113436432A (en) | Method for predicting short-term traffic risk of road section by using road side observation data | |
| CN115588310A (en) | Vehicle collision risk prediction method based on trajectory data | |
| CN113611157B (en) | Method for estimating rear-end collision risk of vehicles on highway | |
| Hyeon et al. | Influence of speed forecasting on the performance of ecological adaptive cruise control | |
| CN116206451A (en) | Intelligent traffic flow data analysis method | |
| CN113761715B (en) | Method for establishing personalized vehicle following model based on Gaussian mixture and hidden Markov | |
| CN112308136B (en) | Driving distraction detection method based on SVM-Adaboost | |
| CN114820216A (en) | Driving risk quantification method, classification method and prediction method | |
| Yang et al. | Dynamic safety estimation of airport pick-up area based on video trajectory data | |
| CN117809458A (en) | Real-time assessment method and system for traffic accident risk | |
| CN112990563A (en) | Real-time prediction method for rear-end collision accident risk of expressway | |
| CN114715146A (en) | Method for predicting severity of potential collision accident | |
| CN114862159B (en) | Evaluation method of automatic driving test scene | |
| CN106570535A (en) | User classification method based on adas (advanced driver assistant system) and obd (on-board diagnostic) data | |
| CN115482662B (en) | Method and system for predicting collision avoidance behavior of driver under dangerous working condition | |
| CN115063976B (en) | Vehicle conflict risk assessment and prediction method based on multichannel convolutional neural network | |
| CN115860461A (en) | Risk factor evaluation method for traffic conflict of non-motor vehicles at plane intersection | |
| CN113252057A (en) | Method and system for identifying driving tendency based on high altitude navigation data | |
| CN113723529A (en) | Traffic information credible identification method based on speed prediction algorithm | |
| Ruan et al. | A novel scoring method for pedestrian automatic emergency braking systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |